本研究以台灣上市櫃公司為研究對象,主要研究目的在建構穩定且具預測力的企業財務危機預警模型。此外,本研究並比較以Cascaded Logistic Regression建構之模型與以Logit Regression模型在使用相同研究變數下模型間的差異。為建立穩定且具預測力的企業財務危機預警模型,本研究在研究年度的選取、樣本配對、配對標準與研究變數的選擇上均經過詳細考慮與檢驗。為驗證本研究模型預測的穩定性,本研究共分為三階段對預警模型進行穩定性驗證。結果發現,即使檢驗時間離模型建構期間不斷增加,模型對於區別出未來發生財務危機的公司依然具有絕佳的預測能力及穩定性。在Cascaded Logistic Regression建構模型與以Logit Regression建構模型的部份,其彼此間的差異並不大,這也顯示出Cascaded Logistic Model在縮減變數的同時,模型並不會降低對正常公司區別率、整體區別正確率、Tpye I Error與模型解釋力。
This paper uses the listed and OTC companies in Taiwan as a sample. The main goal attempts to construct a stable prediction model. In addition, this research compares the differences of using the same variables in Cascaded Logistic Regression and Logit Regression. In order to establish stable and useful financial distress prediction model, we pass through the detailed consideration and validation in the research year selection, the sample pairs, the standard of pairs and the research variables choice. Support Logit Model and three stages of examination are proposed to predict financial distress. In empirical results, our model has excellent forecasting ability and stability following time passing. The final part section compares the differences between Cascaded Logistic Regression and Logistic Regression, its differences between each other aren’t big.